AWSF TAC Meeting - April 10, 2019

Voting member attendance

[x] Daniel Heckenberg - Chairperson, Animal Logic Pty Ltd [x] Gordon Bradley, Autodesk [ ] Mark McGuire, Blue Sky Studios, Inc. [ ] Michael O’Gorman, Cisco Systems Inc. [ ] Henry Vera, Double Negative [x] Bill Ballew, DreamWorks [ ] Matt Kuhlenschmidt, Epic Games, Inc. [x] Brian Cipriano, Google [x] Jim Jeffers, Intel Corporation [x] Larry Gritz, Sony Pictures Imageworks [x] Jean-Francois Panisset, VES Technology Committee [x] Cory Omand, The Walt Disney Studios [x] Kimball Thurston, Weta Digital Limited [ ] Ken Museth, OpenVDB Representative [x] Michael Dolan, OpenColorIO Representative [ ] Eric Enderton, NVIDIA

Apologies

Henry Vera, Double Negative

Other Attendees

  • Andrew Grimberg (Linux Foundation)
  • David Morin (ASWF)
  • Dan Bailey (ILM)
  • John Mertic (Linux Foundation)
  • Cary Phillips (ILM)
  • Francois Chardavoine (ILM)
  • Robert Vinluan (SideFX)
  • Jonathan Stone (Lucasfilm)

Agenda

  • Goals for TAC: Year 1

  • Result of OpenCue adoption vote
    • John: OpenCue has fully passed, we got all the votes.
  • Proposed updates to project lifecycle
    • A few dissenting votes on OpenCue seemed to be related to lifecycle of OpenCue project compared to other projects so far. Should we discuss the concept of the “incubation stage”: that’s where a project goes to ‘get its stuff together’, for other projects it’s been the administrative steps of code repo transfers, licensing, contribution agreements. For OpenCue there’s the added dimension of it being a “new” technology (at least as an open source project). Should there be additions to the project lifecycle document that address these issues. Larry: does the current document speak of a “time limit” for the incubation phase, or should it be driven by the project itself? Or it is the responsibility of the TAC to make determination of end of incubation phase. The documentation speaks of an Annual Review for this case. Should this be updated / clarified? Similarly there is some documentation that speaks of the Graduation Review which doesn’t have specific language about adoption.
    • Timeframes for phases?
    • Henry:
      • specific project criteria for graduation
      • Resource commitments and roadmap to address above?
    • Larry: exit path from incubation
    • John: asking Bill about his concern with
    • Kimball: we should have a discussion about resource prioritization between projects.
    • Daniel: will suggest some additions to the lifecycle document on TAC mailing list, as well as re-emphasize the existence of that document.
  • OpenEXR Adoption
    • Proposal
    • Presented by Jonathan Stone, Lucasfilm advanced development group, lead on Material X, has been taking lead on several Open Source projects at Lucas Film
    • Created in 1999 at ILM, open sourced in 2003, received an Academy Sci-Tech Award in 2007.
    • Believes the incubation stage requirements are already met.
    • Adopted stage requirements:
      • 39 committers on GitHub, individuals and companies
      • No CII badge yet
      • 1279 commits over GitHub lifetime
      • No documentation of list of committers
      • No technical lead appointed
      • No license scan submitted to TAC
    • Near term goals:
      • CMake-based build / test infrastructure on near-term roadmap
      • 23 open pull requests to be merged
      • Need a way to report security issues
      • Should IlmBase be a separate Git module?
      • Discussion on lightweight, shared library for vector/matrix functionality
    • SIGGRAPH 2018 BOF feedback
      • Simple and fast write path (along the lines of TinyEXR)
      • Integer image support (reducing the need for TIFF)
      • Standalone LibHalf library (and reconciliation with CUDA half type)
      • GPU decompression for performance (possibly supporting only a subset of formats)
      • Metadata standardization
    • Proposed list of initial Technical Steering Committee, also looking to give write access for recognized committers
    • Jim: does the current decompression use SIMD / parallelism / AVX512? Kimbal: fully multi-threaded, some snippets of SIMD code, but goal is to be portable, then be fast. So majority of the code is “generic CPU but multi-threaded”. Jim: may be an opportunity to pursue “GPU + SIMD” decompression. Larry: currently multi-threading is only in case where application is reading multiple scanlines at once. Part of the point of GPU decompression is when the end destination of the image is the GPU anyway. Usage patterns make it not attractive to use GPU decompression if you have to read the image back right away.
    • Daniel: project should move from current CLA to Linux Foundation minimal CLA. License for project is a modified 3-clause BSD. John: there are also some CLAs in the TAC repo. Jonathan: looking at adopting stock Linux Foundation CLA.
    • Vote on adoption to incubation stage. Daniel calling for the vote: Bill seconding. Project is approved unanimously.
    • Cary Philips: TAC should be involved in discussing the future of LibHalf / iMath since several facilities have a lot of code based on those libraries, so changes should not be undertaken lightly. Daniel: ASWF is the right place to discuss these impacts. Cary: iMath could be a separate project outside EXR, that’s not really a problem, but changing the API would. Daniel: every project ends up folding in its own math library… Justifiable but creates issues at scale.
  • Technical Project updates
    • OpenVDB
      • Dan Bailey: still looking at CI and CMake as main 2 projects on the roadmap, lots of progress being made. Should be able to migrate to CircleCI in next 2 weeks.
      • Received proposal from Autodesk for multi-resolution grid storage, will be discussing in next meeting, whether good fit with OpenVDB.
    • OpenColorIO
      • Mike Dolan: put out a 1.1.1 release last week, first release in over a year, mostly bugfix / patch release.
      • Working on CII badge requirements, main focus of recent discussion. Some aspects were deferred on wider TAC-wide decisions, in particular security considerations and minimal static analysis requirements. Don’t want to go too far down the road if the wider community will decide to go in different direction. Daniel: lines up with OpenVDB situation, OpenCue will bring very different requirements.
      • Working on license scan issues, having discussions with ACES project over the ACES config that is widely used in OCIO repository and has license differences.
      • Approached by Marvel, interested in contributing to OCIO in near future.
  • CI updates
    • Circle CI
      • Andrew: no progress on discussions with CircleCI folks, meeting still hasn’t happened. Still trying to resolve that issue. From perspective of CI systems, RelEng evaluating Circle CI and Azure Pipelines from wider perspective of several projects under LF.
      • Internal discussion on Gitlab CI as a potential option (you can use GitLab CI against GitLab or GitHub). GitLab CI runner service supports all 3 major platforms.
    • Working group: see notes from last meeting
    • Jim: pointing out the Professional CMake book, which is a good, prescriptive book on how to use CMake.
  • John: CommunityBridge introduction
    • May have been brought up on previous TAC call / email thread
    • Platform debued by Linux Foundation to help sustainability of Open Source projects, for instance mentorship programs.
    • Would TAC be interested in a presentation by someone from the CommunityBridge project? Daniel: yes, we would be interested.
    • John will make connection between team and Daniel, will be presenting at a future TAC meeting.
  • Update on industry / candidate projects
    • DWA: USD Manager and others. Lightweight USD browser. About to contribute Nuke plugin that read / writes USD. Larry: what were considerations of making this a separate project instead of contributing back to USD project itself? Based on feedback from USD group. Gordon: probably same with USD Maya plugin. Larry: thinking about USD as a single, integrated “package” vs multiple independent projects. Message from USD team at SIGGRAPH 2018 was that plugins would move to external repositories. Jim: released USD Hydra / OSPray render as a separate plugin.
    • Data projects (e.g. wikihuman, datasets like Moana Island)
      • Mostly static datasets
      • Different license requirements
      • Is there value for the ASWF in hosting those?
      • JF: could be used for testing infrastructure of existing projects
      • Gordon: job of hosting data could be quite different than hosting software / code. Should we focus on software / code for now? Those discussions came up earlier. Daniel: trying to get a sense as to whether there’s been a shift of opinion. Jim: they do load the Moana dataset every day, so it’s been hugely useful for them, but also understand that it is a different undertaking to host these datasets.
    • Vector / Matrix library
    • Camera raw library
    • others?
  • Next meeting
    • April 24, 2019

Action Items (AIs)

Notes

Chat